Unsupervised Learning via Meta-Learning

TL;DR

This paper introduces an unsupervised meta-learning approach that constructs tasks from unlabeled data to learn a versatile learning algorithm, outperforming previous methods on various downstream classification tasks.

Contribution

It presents a novel unsupervised meta-learning framework that automatically constructs tasks and optimizes for rapid learning, demonstrating significant improvements over prior unsupervised methods.

Findings

Effective task construction via clustering embeddings

Improved downstream classification performance

Applicable across multiple image datasets

Abstract

A central goal of unsupervised learning is to acquire representations from unlabeled data or experience that can be used for more effective learning of downstream tasks from modest amounts of labeled data. Many prior unsupervised learning works aim to do so by developing proxy objectives based on reconstruction, disentanglement, prediction, and other metrics. Instead, we develop an unsupervised meta-learning method that explicitly optimizes for the ability to learn a variety of tasks from small amounts of data. To do so, we construct tasks from unlabeled data in an automatic way and run meta-learning over the constructed tasks. Surprisingly, we find that, when integrated with meta-learning, relatively simple task construction mechanisms, such as clustering embeddings, lead to good performance on a variety of downstream, human-specified tasks. Our experiments across four image datasets indicate that our unsupervised meta-learning approach acquires a learning algorithm without any labeled data that is applicable to a wide range of downstream classification tasks, improving upon the embedding learned by four prior unsupervised learning methods.